Hi Mike,
You are not alone in the uphill battle of selling a SAN to management. Management has a hard time with soft-dollar benefits like making you more productive and reducing the after-hours work you currently have to do. Sometimes the high availability and disaster recovery benefits are more tangible to management, so give that angle a try. The entry-level SAN boxes from most of the hardware players come up short on performance, features, scalability, and redundancy. Hardware companies do not want to cannibalize sales of their midsize offerings with their entry-level offerings.
Quick review of storage performance: All sustained storage performance is ultimately gated by the number and performance of disk drives in the solution. Weak controller designs will prevent you from obtaining the full potential of the drives contained in a solution, but the fastest controller in the world can't perform better than the drives behind it. The same is true for the VSA. The VSA delivers roughly 80-90% of physical SAN performance per spindle. So if you build a VSA configuration with two ESX servers and two VSA licenses and each ESX server has 6 SAS disks, you will get 12 spindles of performance. These 12 spindles will deliver 80-90% of the performance of 12 SAS spindles running in one of our physical SAN units. Our physical SAN units are among the best-performing enterprise-class SANs on the market today. We outperform high-end SANs from the big guys on a daily basis.
The VSA can indeed present iSCSI volumes to other, non-ESX servers as well, so your physical MS SQL 2005 server can access a volume that lives on the VSA.
There is a HUGE difference between entry-level hardware SANs and a VSA-based SAN in terms of scalability. Let's imagine that you have two ESX servers in your main location and each one is running the VSA. Each of the servers has 6x 300GB 15k RPM SAS spindles. Your configuration will deliver some measureable amount of performance - say 2000 IOPS. If you add a third ESX server, you are adding CPU and memory resources to share in your VMware cluster AND you are adding storage resources to your VSA cluster. You will now have on the order of 3000 IOPS of disk performance. Adding six more SAS drives to your twelve-drive VSA storage cluster does not incur any downtime, and you now have 18x 300GB 15k RPM SAS drives serving I/O for ALL volumes on the virtual SAN (we spread the data from existing volumes onto the new spindles automagically). Note that many of the "entry-level" physical SANs can't scale beyond 12 drives and often are limited to SATA disks only. They also introduce single points of failure - the SAN infrastructure components (switches, HBAs, cabling) and the SAN itself (controller chassis, midplane, disk shelf) become liabilities. With the VSA, the three-ESX-server-with-VSA configuration described above can lose 6 disks (one entire system) along with a drive in each of the survivng units (RAID 5). Now that's redundancy.
See the InfoStor review for an indepent look at the VSA:
http://www.infostor.com/display_article/314533/23/ARTCL/none/none/1/SANitize-DAS-to-enhance-virtual-...
All of this begs the question - when do you choose a LeftHand Networks physical SAN over the VSA. Capacity and hardware management are the answers.
Capacity: The VSA can manage 2TB per license. As you can seen in the example above, it is easy for a VSA SAN to scale in 2TB increments. If you need 10TB of space and you want to be able to sustain some system failures, you're going to need 10 ESX servers with 2TB each. Not necessarily ideal when you consider than LeftHand Networks can do that with SATA disks in a two-node physical SAN configuration today.
Hardware management: The disks in the ESX server running VSA are managed by the ESX server (or the hardware RAID controller supported by ESX). Any drive failure alerts need to be managed at the ESX level - the VSA has no visibility into the physical hardware layer (obviously, since it is running as a guest OS). A physical SAN from LeftHand Networks does bad block remapping and defragmentation and other drive-health monitoring (SMART for example) as part of our self-healing SAN feature set. Add phone home support and the ability to scale storage capacity and performance independently of ESX performance, and you see the benefits of a physical SAN make sense for larger companies.
Interestingly, we have customers with big, legacy, fibre-channel SANs running the VSA on their ESX clusters to create another storage tier. Pretty cool way to use the disks that are already spinning in an ESX cluster - you're paying for the power and cooling of those drives, so why not store something there? Without the VSA, those the data on those disks could be lost - there is not fault tolerance. With the VSA, they have a scalable tier of disk that is fault tolerant.
I hope this helps - the best advice I can give you is to download the VSA and try it out - the demo version runs on VMware Server or ESX. After the 30-day evaluation period, it will continue to work as a simple iSCSI target until the end of time.
Feel free to PM me if I can be of assistance.
Happy New Year!
Stephan Stelter, LeftHand Networks
